Research, Risk Assessment Techniques Based On Bayesian Network Software Requirements

Risk Evaluation on dependable software is a core step of risk management, and software reliability is a key attribute of dependable software, so it has important significance for improving software dependability to research on risk evaluation of reliability. In variousstages of software development——requirements analysis,system design,coding and testing,there is a large number of risks. If the risks can be identified early in the process of softwaredevelopment——requirements analysis phase, software reliability can be improved to amaximum extent. In view of bayesian network is effective for risk evaluation, the focus of this dissertation is to establish a software requirements-risk evaluation model based on bayesian network. By studying risk Bayesian network models,structure and probability inference method, provides a systematic modeling method for software requirements-risk evaluation. The main content and fruits of this paper are outlined as follows:First of all, on the basis of study on concept of dependable software,attributes of dependability and process of software development, carrying out research on risks that may occur during requirements analysis phase, identifying the main requirements-risks, and making a risk list.Secondly, to research on risk evaluation of software project, uncertainty contained in the main activities of requirements analysis phase should be analyzed one by one, so that sources of risks related to every requirements-risk in the risk list can be determined.Then, a risk evaluation model based on bayesian network is established. The dissertation describes the design principles and evaluation process, refines the sub process of the model, proposes structure-learning algorithm and inference method, and also uses ahp for evaluation of risk effect.Finally, a case of software development is showed to verify the model. By identifying the risks and sources, construct corresponding model; using product of risk probability and risk impact to quantify the result of risk assessment, to evaluate reliability of the software, which help the manager and developers recognize high-risk modules, make risk management much easier and more effective, to enhance the software reliability.